AI RESEARCH

Omni-Video: Democratizing Unified Video Understanding and Generation

arXiv CS.CV

ArXi:2507.06119v4 Announce Type: replace Notable breakthroughs in unified understanding and generation modeling have led to remarkable advancements in image understanding, reasoning, production and editing, yet current foundational models predominantly focus on processing images, creating a gap in the development of unified models for video understanding and generation. This report presents Omni-Video, an efficient and effective unified framework for video understanding, generation, as well as instruction-based editing.